Episode 237: FF-P-MEG

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Podcasting 2.0 for October 10th, 2025, episode
237, FFPMeg.
Hey everybody, it's good to be back.
Are you kidding me now?
Two weeks and so much has already happened.
Welcome to Podcasting 2.0, the official board
meeting of all things podcasting.

(00:21):
If it's going on in podcasting, it's talked
about here.
We are in fact the only boardroom that
gets none of the $3 billion in projected
advertising.
I'm Adam Curry here in the heart of
the Texas Hill Country and in Alabama, the
man who's every man's philosopher and an agent
of change.
Say hello to my friend on the other
end, the one, the only, Mr. Dave Jones.

(00:48):
I guess it's because I don't have FFMpeg
installed.
You need FFMpeg.
Everybody in the world needs FFMpeg.
Well, I just, I mean, I figured it
would just be there.
I mean, I couldn't build a robot out
of FFMpeg.
FFMpeg, sudo apt-get.
Install.

(01:10):
Install FFMpeg.
Unable to locate package FFMpeg.
Well, that's a problem.
You forgot to do a sudo apt update.
How could you not?
Like, how could this, this has to be
right.
I actually had to build FFMpeg from source
to get the RTMP module.

(01:31):
Now, did you have to do this?
Oh, yes.
Or did Grok just tell you you had
to do it?
No.
No.
After I found out that it did not
include the RTMP module, I went to looking,
but it's always interesting then, because, you know,
it might install it in a different directory
than you expect it.

(01:51):
And if you don't explicitly tell Grok what's
going on, you can go down a rabbit
hole for an hour or two.
Sudo apt-get install FFMpeg does not, does
not work.
Why is this?
I don't understand.
Where is it?

(02:12):
Where is it?
It says no, unable to locate package FFMpeg.
What is my problem?
You got no package, man.
Your package is no good.
I've heard that, but now I know for
sure.
You got no package.
It's weak.
Let's see, cat.
Okay, cat etc message.

(02:33):
Oh, Eric.
Ericpp is trying to be helpful.
Nix-shell.
Okay, yeah.
Now, we're not all running on Nix.
Oh, Nix, oh, yeah.
No, that's not.
I'm on like Ubuntu something.
I don't even remember what version.
Ericpp, you may not have a spouse or
a girlfriend.
If you're on Nix OS, then I don't

(02:54):
think you can have a relationship with anything
else.
Oh, I'm on Ubuntu 20.
I may be off.
Oh, no.
I may be off long-term support and
I can't install packages.
Is that really what this is?
I guess so.
If it is, then I'm gonna be really
annoyed.
Did you kill your stream?
Well, it killed itself.

(03:15):
Oh.
It gave me an encoder error and just
killed it and just died.
Oh, no, that's no good.
That's no good.
Hello, everybody.
Welcome to Podcasting 2.0, where we take
you to the command line, every single show,
just to let you know that we have
no idea what's going on in the world.
I certainly have no idea what's going on

(03:35):
right now.
Is it Snap?
Am I gonna have to resort to Snap?
Let's bring in the memory leak that is
known as Snap.
I'll tell you what, while you're doing that,
why don't we bring in our special guest
for today?
Because he knows a lot more than just
what he's here to talk about, which is,
of course, shameless self-promotion of a product.

(03:55):
But that's what most people should be doing
here.
Please welcome the owner, creator, the proprietor of
PodChapters and many more, Daniel J.
Lewis.
Hey, guys.
Hey, Dave, try and stick some JSON in
it.
Ah, there it is, there it is, there
it is.
Sue, your comments are not helpful, Daniel.

(04:21):
Pseudo apt, this is horrible.
This is horrible.
Pseudo apt again.
I'm just gonna try, wait, unzip.
Let's try to, okay, that works.
So it's not a package problem.
Yeah, that's what she said.
The package is fully operational.
Your package is fully operational and loaded.

(04:42):
Yes, it's ready to fire.
But this is not, but the, is it
caught?
Oh, I see what my problem is.
I'm such an idiot.
Okay, now we all want to know.
Now we all want to know.
Misspelled?
I misspelled FFPEG.
Ah!

(05:04):
I typed FFPEG.
Oh, there you go.
FFPEG, okay, for anybody who's won, FFPEG, FFPEG.
For anybody who is curious, FFPEG is not
a real package.
But it's a show title for sure.
FFPEG.
FFPEG, oh brother.

(05:25):
That sounds like a concealed carry kind of
thing.
It's almost lewd, but in a way that
I can't fully understand.
Well, snap, I had to use snap though,
which is also not good.
So what we're doing here is, for those
who are wondering, for the podcaster, godcaster, I'm

(05:47):
sorry, we have, I don't know how many
hundred stations we have.
And we create these channel streams.
So you throw a bunch of podcasts in
a player and then it plays the most
recent episode of whatever's most recent and does
that in a loop with station IDs in
between.
And we've enabled a way to take over

(06:09):
that stream and go live.
IceCast was pretty simple.
Most churches use RTMP.
So they're streaming video to exciting platforms like
YouTube and Facebook.
So exciting.
And since we refuse to be streaming video,

(06:31):
we have a little RTMP to IceCast bridge
and it actually works.
Except for the metadata.
Metadata is turning out to be a bit
of a problem, not because of the RTMP
to IceCast or how we are doing metadata.
Okay, it seems to be working.

(06:53):
Yeah, hold on, I can see it rolling.
You do see it rolling?
Well, I see you're being probed.
Yes, your stream is active.
Player eight, player eight, player eight enters the
game.
Scanning.
Okay, it has not crashed on me on
the output yet.

(07:13):
Is it actually making a file?
But this is all useless anyway, because I
didn't start at the right time.
So it's going to be way off again.
Oh, that'll be fun.
Let me see, I can check and see
if we're live here.
It freaks out all the apps.
Is it actually making a file?
This is crazy.
There you go, you're live, okay, you're live,
awesome.
It works.
Well, okay, it's creating .ts files.

(07:35):
See what I'm trying to do is, like
a good computer guy, I changed the settings
on OBS right before the show.
Oh, of course, yeah.
Like I connected to the stream, I heard
Daniel's voice and I hit settings.
And I went in there and just started

(07:56):
twiddling with stuff, because what I want to
do is I want to record directly to
HLS instead of having to re-encode afterwards.
That way I can just go, shoot it
right up to Linode.
We had the funniest episode with VoidZero.
On Wednesday, he sends me a note, which
I will read verbatim just because it's funny.

(08:17):
No agenda?
Yeah, for no agenda.
He says, fixes changes and updates today.
Any report of the stream being offline?
I'm aware of it, migrating a bunch of
stuff.
People won't notice.
Yeah, everything broke.
The stream broke, the website didn't update.

(08:41):
Wow.
But it's like, you know what's going to
happen.
It's like, just don't say no one will
notice because there's always something you notice.
It always breaks.
It's hilarious.
I thought Martin knew about no change Thursdays
and Sundays.
Well, he did it on Wednesdays, to be
fair.

(09:02):
That was the good part, but I'm like,
no, just don't do it.
Don't do it, man, and he did it.
Do you make changes?
There are some people in the universe who
have the ability to make Mercury go in
retrograde.
Precisely.
Daniel, do you change critical parts of the
infrastructure of PyChapters like on Friday night?

(09:24):
No way.
No better than that.
No, that's not a good idea.
I, however, still have a tendency to do
that.
Oh, you just pushed to production, that's you.
Oh, I go, yeah, I'm hacking on production.
Staging is for pussies, absolutely.
That's what it is.

(09:46):
You're a real man.
Well, I had the wonderful experience of a
dev server that got, I logged into the
dev server.
I'm like, why is it, why am I,
is everything in 100 CPU cycles?
What is, hmm, so.
Oh, yeah, yeah, your Bitcoin miner expedition.

(10:06):
So it happened to three machines before I
could batten down the hatches, and then somehow
one of the machines I thought I had
closed up, I guess that somebody left something
in there, and it may not have been
a Bitcoin miner because I got a Linode
strike.
Linode, so you get like, I guess, Akamai

(10:30):
complained that my server was abusive.
I'm like, bad server, bad server.
Oh, your server was trying to hack other
servers.
Pwned for something else, yeah, but it was
still the same kaudit zero process that was
running, but it was hidden in all kinds
of places.
It had all kinds of scripts that would
restart it, reinstall it.

(10:52):
It was fun to track down.
Well, I mean, just, I would never, if
a server, if any machine ever gets popped
like that.
Oh, no, I- Never trust you, you
have to rebuild from scratch.
I did, I did.
I rebuilt from scratch, which gave me the
opportunity to rebuild it on a smaller Linode,

(11:14):
which is kind of fun, but yeah.
Linode give you a strike against your VIA
or against your account.
Well, no, not a strike, but they lock
down your networking, so you can't get into
it.
So they just lock you down.
Well, that's kind of cool, though.
Yeah, well, you can get into it with
Lish.
Lish.
Yeah, Lish, but you can't remote into it,
right?
You can't, no, no, no, no.

(11:35):
Lish or Glish, your choice, Lish or Glish.
It is, your choice, whatever you want, or
fish.
Hey, Daniel J.
Lewis, it's good to have you in the
board meeting.
It's been a hot minute.
Yeah, been busy building things, talking to people,
trying to still spread the gospel of Podcasting
2.0 as well.
Well, let's talk about your pod chapters, because
I got an early beta test, which I

(11:57):
thought was really phenomenal.
I went back to use it again, and
then it said, well, you have to become
a beta tester, and it was all kinds
of confusion, so I figured I'd just wait
until I can pay for it, because I
want to pay for it, because it is
that good.
And it's a sneaky way, kind of, of
you trying to build your super chapters idea.

(12:17):
I see what you're doing.
Yeah.
Why don't you just call it super chapters?
You had to throw a pod in front
of it.
Unnecessary roughness on the field.
So tell us about it, because I saw
this, I'm like, yeah, this is good.
This is exactly what we need.
Tell me, tell us what it is, and
I want to know how the adoption rate

(12:39):
is.
Yeah, I have been using chapters for a
long time, and I followed the legacy chapters
for years, and also used Podcasting 2.0
chapters.
You mean heritage chapters?
I prefer legacy.
Vintage.
Vintage, ooh, that's a good one, yeah.
Although vintage implies something of high value.

(13:00):
But anyway, I like supporting the many different
formats, because I still use a podcast app,
Overcast, that does not support Podcasting 2.0
features, unfortunately, and I know many other apps
still support the legacy, vintage, whatever, old-style
chapters.
And the chapter experience was so frustrating, because

(13:21):
for one thing, there's just not a good
cross-platform solution to be able to recommend.
There's almost no way to make chapters on
Windows, and the best app that was available
for Mac that did support Podcasting 2.0
had all kinds of weird issues, and it
was just a pain to work with.
What app was that?
It's called Podcast Chapters.

(13:42):
Oh, gosh, that's been around forever, hasn't it?
Yeah, I used to use that.
I think maybe 15 years ago, I was
still using that.
Maybe I'm wrong.
Yeah, that was the initial JSON format that
we used.
We kind of merged what we were working
on with what he was working, with what

(14:03):
he had, and came to this sort of
amalgam.
And so, yeah, he was one of the
first people to support it.
Yeah, and it worked.
It just was so cumbersome to work with.
And then, as I started getting into certain
things, there were several companies that I talked
to that use AI and can add chapters,
but most of those chapter-adding AI models

(14:25):
just create way too many chapters.
Like, by now, we probably would have created
10 chapters with some of these tools, because
of how many topics we've talked about.
And I wanted to use AI to place
my own chapters, because when I record, I'm
talking from an outline, so I know exactly
what I want my chapters to be.
I just don't know where I want them

(14:46):
to be.
And so, this workflow that I had was
using three or four different apps and taking
15 to 30 minutes for every episode just
to add the chapters and transcript and do
some of this stuff.
So I just decided, I just want to
make a tool that works better for me
to do all of this.
And as I'm making it, I realized someone

(15:07):
else might like this too.
I'm going to turn this into a product.
And thus, PodChapters was born and launched in
about two months.
All right, so tell us exactly what it
does, because it does a lot of things,
and it's actually quite impressive.
You could almost have a basic podcast host
and add PodChapters and have a whole bunch

(15:27):
of extra features.
Oh, yeah.
So some of this ties into the difference
between chapters and what we're calling super chapters.
I like that Dovi Das calls them super
chapters, so I call that something else.
So this does kind of tie in and
bridge with that.
But a chapter, just in case anyone's not
familiar, a chapter, whether that's legacy chapters or
Podcasting 2.0 chapters, they're functionally the same.

(15:50):
They are a time within the audio with
a title and an optional link and an
optional image.
Podcasting 2.0 chapters add on the location
to that and the ability to hide something
from the table of contents.
But still, they're functionally the same.
And so what PodChapters does is you upload
your audio to it, and there will be

(16:11):
an option in the future that you'll be
able to upload video to chapter your videos.
But for now, it's just MP3 audio.
You upload your audio to it, and you
can either BYOT, bring your own transcript, or
you can have it generate a transcript for
you.
And the transcript generation is blazingly fast.
And I've built in some things that really

(16:31):
help it too, where you can tell it
a list of proper nouns or key terms
so that when it finds something that is
close to that, it can transcribe it properly,
like John C.
Dvorak.
It can finally get that correct.
And you can also give it- Wait,
how did you do that?
How did you get, I have found very

(16:53):
few AIs that can consistently recognize the name,
the spoken name Dvorak, and get it right?
It's part of the AI protocol that I'm
working with, which is DeepGram.
They have a really good model for transcription.
And in their code that you can work
with, you can send it these key terms

(17:14):
so that when it gets to something that
it thinks is close enough, and it's not
always 100% accurate, but it is very
accurate and very good at this.
And it can even do find and replace
options on top of that.
Like for example, frequently when I speak a
URL in my own podcast, like slash give
back is my own value for value page.

(17:34):
It sometimes transcribes it as slash give space
back.
So I just put in there a find
and replace.
So anytimes it sees slash give space back,
replace that with slash give back, no space
in there.
And that kind of thing saves a whole
lot of time too.
So when it uses the transcript, the whole
reason why it needs a transcript, you either

(17:56):
bring your own SRT or VTT, or it
generates one for you, is so that then
it can find where chapters could be.
Then you have the choice of, do you
want to either give it your own outline
to create chapters from your outline, or do
you want it to suggest chapters for you?
Both ways work with the transcripts so it
finds where that information should be.

(18:19):
And it does a really good job of
even suggesting the outline that you already speak
into it.
I've fine-tuned this thing like crazy to
make it do that.
And then it adds those chapters both to
the MP3 file, as well as gives them
to you in multiple formats, the JSON format
for Podcasting 2.0, Podlove simple chapters format,

(18:40):
it gives you YouTube timestamps, it gives you
an outline with timestamps, and it can even,
if you're on a hosting provider that you
can't upload or work with their own chapters,
it can even host the chapters for you
at a predictable URL based on the file
name of your episode.
So you could know, like in your case,

(19:02):
Adam, I was thinking specifically about what you
do with NoAgenda in this case, and Dreb
Scott does for the chapters where you don't
have the chapters ready when you publish the
episode, but you want to have a predictable
URL.
So that's what this can do is that
it can create that URL that has a
unique stream in it, but it doesn't change
per episode.
The only thing that changes per episode is

(19:24):
the episode file name.
So then it can host the chapters for
you, and by hosting the chapters for you
in the future, when chapters can start doing
some cooler, smarter things, I'll be able to
build that into the system by providing chapter
hosting as a service, or people can download
it and host them themselves if they want.
So what's the uptake?

(19:44):
How long has it officially been open and
available?
It's only been out for a week.
I launched it on International Podcast Day, September
30th.
Oh, damn, I miss it again.
International Podcast Day.
What day is that?
I need to put that in my calendar.
September 30th.
September, who came up with that?
Why is that a?

(20:05):
I was one of the co-founders of
that.
It was Steve Lee, Dave Lee.
That's why you remember.
Okay, I got it, yeah.
Yeah, I celebrate it more than I celebrate
my own birthday.
Okay, and so I did hear about it
in pod news, but can we get a
presidential proclamation or something so it's a little

(20:25):
more solidified in the books?
At this point, it's pretty close to that
because it was 2014, I believe, or 2013,
that Steve Lee, Dave Lee approached me and
Dave Jackson and a couple others to say,
hey, how about we try and start this?
And one of the reasons why we picked

(20:47):
September 30th, it's easy to remember, last day
of the month of September, but also there
are a lot of significant things about podcasting
that happen around September, including your birthday, Adam.
Wow.
It was one of the things in consideration.
The beginning of the month, but okay, yeah.
But also certain things like the history of
podcasting where the first podcast, at that time,

(21:11):
our understanding of the first podcast was published
around September and also certain other significant things.
So there were a lot of, it's kind
of like President's Day.
There isn't one single president that was born
on President's Day.
It's representing multiple things around the same time.
But can we get a mattress sale on
International Podcast Day?
That's what we need.

(21:32):
Sheets, sheets and linens.
Linens, yeah.
The YSL.
Fireworks would be great.
Yeah.
Make them all in the RSS icon.
So what are you using for the AI
stuff?
And this is, they were talking about this
on Pod News Weekly Review with the guy

(21:55):
who's doing a video-first hosting company.
He has a lot of AI stuff in
there.
I mean, how do you budget, basically?
How do you budget for the usage, for
the credits, all that stuff?
It seems like it's kind of a crapshoot.
Yeah, that was a real challenge.
And for anyone doing anything with AI, that

(22:17):
is going to be the biggest thing that
you'll have to think about because you really
don't want to offer unlimited unless you can
completely predict what your output sizes will be
and how the models will work.
But you really, I think you have to
make it some kind of credit system.
And that's what I went with is charging

(22:38):
based on AI credits where each level that
you subscribe to, you get a certain number
of AI credits per month.
I originally played with calling them transcription minutes
as a type of currency within the app.
But then I realized not only is that
kind of hard to understand in certain contexts,
but it's also very long, a lot of
letters, transcription minutes.

(22:59):
So I just went with AI credits.
And people have to sometimes just look at
how much is it costing to generate this,
do a lot of tests, figure out what
are the edge cases, see how much it
actually costs you depending on the models that
you want to use.
Like, for example, CloudSummit is one of the
most expensive models.

(23:20):
I would say it's the most expensive popular
model.
There are cheaper models like GPT-5 is
cheaper and GROK-4 is cheaper.
Gemini 2.5 is close to the same
price as GROK or as a GPT-5.
But some of these have different versions of
the model too.
Like you might have a GPT-5 regular,

(23:44):
GPT-5 chat, GPT-5 mini, GPT-5
nano.
And for each of these, they come at
different costs and produce different results, different qualities.
And that's where you have to just test
it, figure out what works.
I made a spreadsheet to help me figure
out what can I afford?
How can this be profitable to me as

(24:04):
well as affordable to others?
CloudMax, let's see.
CloudMax from $100.
Wow, you started a hundred bucks?
That's from Max.
Oh, that's, you're looking at directly the subscription
from the AI provider.
Yeah, like the individual pricing.

(24:25):
Yeah, I'm just looking at this.
API pricing, I'm sure is way more complicated.
If you're looking at integrating AI into an
app, look at OpenRouter.
OpenRouter, oh, I've heard of OpenRouter.
Yeah, OpenRouter.com has all of these models.
And so you pay based on usage and
the usage pricing is the same as if

(24:46):
you were using the API directly from the
companies.
But what's great is sometimes you can get
faster service because they negotiate with other providers
and such.
But you can also very easily switch different
AI models completely without having to go and
change API keys and sign up for somewhere

(25:06):
else and add all of your building.
You can just switch which model you're using
right with your OpenRouter calls.
It's fantastic for you working with AI.
A million free BYO key request.
Okay, so when you use this platform, you're
generating keys on all the other APIs and

(25:27):
then you're just sticking your keys in here
and then you're calling this and it calls
them for you.
Not quite, their BYOK is a special promotion
right now for bring your own key.
The default way that it works is they
give you one API key and you send
your requests to them and each of your
requests include what model that you want to

(25:48):
use like OpenAI slash GPT-5 if you
want to use GPT-5.
And then it responds back in a same
format for all of their models respond back
in the same format.
They accept the same input format.
So it's really great for, this is what
a lot of the multi LLM chat platforms

(26:11):
like Magi or T3 chat and other things
like that that make it so easy to
be able to switch between different LLMs so
you can see which one really works the
best.
And like, that's what I do with pod
chapters where I've tested a bunch of the
chapter placement and suggestion on different models and
found that this one works really good.
This one works a little bit better and

(26:32):
this one seems to work the best and
they tend to line up with the cost
too.
So in pod chapters, I have a good,
better, best selection where most of the time
people can be served just great with the
good option, but they can bump it up
higher.
That does cost them more credits but it
might have the possibility of producing better results.

(26:54):
Interesting.
This is a, yeah, I don't know what.
You sound exasperated, Dave.
No, no, that's, I just finished, I just
finished reading the Paul King's North book against

(27:15):
the machine.
And I'm so just down on where everything
is going.
I'm one step away from depressed.
You're depressed.
You need some pharmaceuticals, man.
I'm one step away from just like depression
about the state of all this.
What's the depressing point about it?

(27:38):
It's, I mean, I was reading this, the
most recent thing, this most recent, like not
defense, but like the most recent thread going
through all of the AI talk is stemming
a lot from, I think there was an

(27:58):
article written about it.
And then Jeff Bezos also had this talk
in Italy that people are referencing all the
time.
And it goes something like this.
The AI boom, so the narrative has changed
completely.
So it's like, now it's not that there's
no AI boom bubble.

(28:21):
I'm sorry, there's no AI bubble.
It's now, yes, there is an AI bubble
and it's going to pop, but it's not
an economic bubble.
It's an industrial bubble.
So the new way that it's being explained
is, and you will hear this if you

(28:43):
haven't already, is remember the railroad bubble of
the 1870s.
Oh, I remember it well.
I remember me and Tina were hanging out.
Yes, just yesterday, yes.
And then remember, and how, yes, it was
a bubble and yes, it blew up and
caused economic harm.

(29:04):
But after it was over, we were left
with all these rail lines and that's great.
And so later these rail lines got picked
up and used and this kind of, we
were left with usable infrastructure after the fact.
But we'll be left with lots of data
centers to mine Bitcoin with.
I'm totally okay with that.
And then the next example is that it

(29:26):
was the same, now they're rewriting history that
it was the .com, same with the .com
boom, the same with the .com bubble, that
yeah, it was a bubble and yes, it
hurt us economically.
But at the end of it, we were
left with all this dark fiber.
Yeah, that's true.
And that's great, and that's great.
And now people are picking up this dark
fiber and lighting it up.

(29:48):
And now look at all the benefit.
So this economic argument goes something like this.
Don't worry about the trillions of dollars of
economic waste, of setting trillions of dollars on
fire, because at the end of it, we'll

(30:10):
get a few billion of usable infrastructure.
I mean, that is the broken window fallacy
taken to the largest extreme possible.
And I was reading an article this morning
that said that the amount of money being

(30:31):
spent on AI infrastructure right now is equal
to, on an annual basis, is equal to
the GDP of Singapore.
And the amount of revenue coming in is
equal to the GDP of Somalia.

(30:55):
And do the delta between those two countries.
And let's see.
Yeah, that's a problem.
Yeah, that's a real problem.
I mean, there's other reasons that I'm semi
-depressed.
I'm not depressed about pod chapters.
I love your product, Daniel.
I wanna be clear.
It's cool.
I'm just like, I love the chapters.

(31:18):
I'm just like, AI, I'm so afraid that
this AI thing is going to be, I'm
afraid that it's going to be a collapse
of monumental proportion.
Well, there are two basic camps of AI
usage.
There's analysis and there's creation.
We can put quotation marks around that, really

(31:40):
big quotation marks around that.
Where like what I'm doing with AI is
analysis.
And that's what I love so much because
there are certain things I've wanted to do
for years with my software products that I
couldn't do because it meant trying to build
my own machine learning algorithms and all of
this stuff.
Now, AI can make that so much better
where I just tell it, this is the

(32:01):
input.
Make sure you give me this structured output
back.
And yes, it gives me JSON back.
I tell it, speak to me in JSON,
baby.
And it does.
And that is great.
I think that will continue and it opens
up all new kinds of possibilities for development.
What I don't like is the creation aspect.

(32:22):
And I think that's what you're talking about.
The implosion, the bubble is people creating content,
creating slop with AI.
That, yes, I'd love to see that bubble
burst.
That's the only product.
That's the only product that people actually want.
Yeah, everything's chatbots.
You know, like the most recent episode of

(32:43):
Pod News, the guy that was from the
video hosting company, he's like, Sam's like, what
do you, what'd you put, how are you
using AI?
And he's like, well, we made a chatbot
that you can ask it about your stats.
You know, hey, which episode got the most
hit, you know, got the most views or
whatever.
Like, why do you need a chatbot for
that?
It's like every hosting company shows you that

(33:04):
data.
You don't need a chatbot for that.
This is the amount, like at least three
times this week, I have asked, I have
asked, and not on purpose, because now these
search results, you know, they just give you
the AI summary, but I've asked three different

(33:27):
questions and it's been blatantly wrong in all
three.
What kind of questions?
And I had to go- What kind
of questions?
Well, the one that's the most recent, the
one from yesterday was, what's the, I'm building
a truck in my backyard.
Oh, okay.
No, I am.

(33:48):
I'm building a truck in my backyard, where
do I start?
That wasn't the question, but that's just a
statement of fact I am building a truck
in my backyard.
So I bought a 76 Ford F-250
four-wheel drive frame a couple of months
ago.
Wow.
And then I bought a, I sold my

(34:10):
65.
So I sold the 65 and I'm using
the money that I got from the 65
to fund the building of this new truck.
And so I bought the 76 four-wheel
drive frame with a drive train and everything.
So then I bought a 72 truck that
I'm going to take the body off of

(34:31):
and do a body swap onto that 76.
So in the end of it, it's going
to be a 76 factory four-wheel drive.
So I'm building this.
This has been, this is an ongoing project.
So I'm building this.
And so what I needed to know was,
is the engine that's in the four by
four frame, is it a 390 or a
360?

(34:51):
And so the Ford three, the Ford FE
engine came in multiple displacements from 352 all
the way up to a 428.
And the 360 and the 390 were the
two most popular ones.
Now the 390 was an option.
And the only way to tell the difference,
because you can't tell from casting numbers, you

(35:11):
can't tell from any sort of external indicator.
The only way to tell the difference between
a 360 and a 390 is by the
stroke.
So you have to put a dowel rod
down into the cylinder, rotate the crank to

(35:31):
where the piston is all the way at
the bottom of its travel.
This is great.
Mark the, yeah, yeah.
Mark the dowel rod, then turn the crank
to where it's all the way at the
top of its travel, then do another mark
and then measure.
Mark the dowel rod, Mr. Cawson.
So a 360 has a three and a
half inch stroke and a 390 has a

(35:53):
3.78 inch stroke.
And I knew that they were different.
I just didn't remember which were the two
numbers.
This is a pretty esoteric question, though, that
you, I mean, this is, this is.
But for Ford guys, it's common.
It's very, everybody knows this.
Everybody, you know, you just may not have
the numbers at the top of your head,

(36:14):
but you know that the only, because the
cylinder diameters are the same.
Everything's the same, it's just got a different
crank.
So it has a different stroke length.
And that gives you the extra 30 CCs
of displacement.
So when you, so I just, I'm like,
okay, Ford 360, 390 stroke length difference.

(36:35):
I stuck that into Google.
And I'm gonna, I'm looking, I'm expecting to
go and like, you know, hit a Ford
forum, click through, and somebody has the information
there.
Well, I get an AI summary at the
top that says, the Ford 360 and 390
engines share the same stroke length.
And I'm like, no, they don't.

(36:56):
And it gives references there to articles.
And you click on those articles and they
both don't say that.
They both say, they give you the correct
information if you click through to something a
human actually wrote.
But if you read the summary, it just,
it just lies.
And I'm like, we're paying billions and billions

(37:20):
of dollars for this thing.
And it is worse than the thing we
had before.
This is, it's hard not to be-
Skeptical?
I'm saying depressed, honestly.
It's hard not to be depressed about the
amount of, the amount of like energy that's

(37:43):
being put into pumping this stuff into every
single aspect of our life.
And it's not, whether the accuracy of it

(38:03):
is almost like, okay, here's my sense, all
right.
We've gotten to a culture now, I'm sorry,
I've gone into full blown rant mode.
Yeah, you, yeah, it's okay though.
Let it all out, let it all out.
Just let it come out, it's fine.
We're here for you.
It's gonna be okay.
Part of this is my simmering back pain

(38:23):
here.
But like, I have to channel this into
the microphone.
But the, so the problem that we've, that
we are encountering across the software industry in
general is that everybody, everyone has adopted, one
of the side effects, one of the many

(38:46):
negative side effects of internet delivery has been
that nobody cares about getting their software right
before they ship it.
So you say, okay, we're gonna get it
close.
You're talking about me?
I'm not, I'm leaving you off the table
for now.
I fit the category.
No, I'm not talking about mistakes.

(39:06):
I'm talking about, you know it's broke and
you ship it anyway.
You know, that's what I'm talking about.
Like bugs are always a thing and they've
always existed.
That's just software.
Software has bugs and you gotta, and sometimes
you have to ship a fix or a
patch.
That's always been the case and that's not
a problem.

(39:27):
The mindset has shifted now to where instead
of shipping a product that you think doesn't
have any bugs and finding out later that
it does and having to fix it, now
you say, oh, well, we know, we have
quote known issues.
Yeah, exactly.
You know that this thing is broken and

(39:49):
you ship it anyway and you're like, because
you need to hit a ship date.
And so then you say, well, we'll just
fix it later.
We'll ship a patch later because it's so
easy.
You know, all we gotta do is just
push this code and everybody just refreshes their
browser and it's great.
And so that mindset has leaked into, I
think that has leaked into AI.

(40:11):
I think the whole mindset of AI is
we know that it's broken.
We know that it gives you false information
a large percentage of the time.
Now, we don't know what that percentage is.
It could be 15% or it could
be 40%.
We don't know.

(40:31):
And it's probably always changing because every time
these models are retrained, they get new things
right and old things wrong.
And so there is a widespread knowledge.
Everyone understands that this stuff is giving bad
information a large percentage of the time.
But it's like, well, we just don't care.

(40:51):
We'll just fix it later somehow, I think,
maybe, perhaps.
And all we need is just a few
more billion dollars.
Now, I think it's a trillion is the
latest I've heard.
Yeah, a few more trillion.
Yeah.
I didn't, this- I'm just worried that,
I'm worried that everybody's building things, and this

(41:13):
will be for you, Dan, but also for
myself, that people are building things that rely
on AI credits.
And one day someone's going to say, okay,
it works, but now you really got to
pay the cost.
And it's going to be significantly higher than
what we're paying now.
Yeah, I've thought about that and have been

(41:33):
concerned as well.
But I think that as the models are
developing more, the possibility to run them on
your own hardware is also developing more.
So even if the cost goes up to
using an API from a place like OpenRouter
or something, I might be able to lease
my own server with my own GPU in

(41:53):
it that could run the model for me.
And then I have that own cost, but
then I have to manage things like queuing
the requests and handling simultaneous things.
But I think it's possible.
Can you set OpenRouter to redirect to something
that you host yourself?
No, but it does use different providers based

(42:17):
on whoever is providing the service the fastest.
So like Gemini, for example, it's not going
to only use Google as the provider, but
there might be some other server farm out
there.
Maybe it's Adam's server, Linode server is helping
to process the request, but it's using whatever
provides the fastest responses.

(42:39):
Okay, so you're telling it, you can put
it into a mode that says, I want
you to use just whatever is currently the
quickest, not necessarily the cheapest.
Well, it will most available.
It's for the model that you want to
use.
So there are lots of providers for that
model.
And if you'd look over at OpenRouter.com

(43:03):
then, or I'm sorry, .ai and look at
their models page, you'll see a bunch of
models that they offer.
They offer hundreds of models.
And for any one of those models, when
you click on it, you can also see
the number of providers.
Many of these models have multiple providers and
they come in at different speeds.
So what OpenRouter does is it routes your

(43:23):
request for that model to whichever model is
performing the fastest.
They do have an auto model that will
route your request to a model that they
think will serve your need the best, but
that also comes at unpredictable costs.
It might go to GPT 4.0, or
it might go to a nano model or

(43:45):
something else.
Yeah, right.
Well, is there a, so you mentioned the
problem and I've seen this before too, where
AI creates like way too many chapters.
Just a chat, like a chapter every three
minutes or something like that.
Oh, that happens a lot, yeah.

(44:05):
Yeah, I mean, how are you controlling that?
That's where my secret sauce comes in, where
I have.
No, it's okay, you can't use proprietary information.
Really, everything you're doing with an AI system
all comes down to the prompt and what
information you're doing with it.
And that's the thing that when I started
building this, I practiced over and over and

(44:26):
over just running the same file through the
prompts, trying to get it to specific results.
And then I would test on other files
too, trying to get it to that point.
And I think I've gotten it there.
And so the systems that produce a lot
of chapters could be using the exact same
LLMs I'm using, but their prompt is different

(44:46):
because when you work with an AI, simply
changing a single word in your prompt could
give you potentially very different results.
Even running the same prompt multiple times can
give you different results.
And some people don't realize that with the
AIs, they think that the AI gives you
an absolute answer and everyone gets that same

(45:06):
answer.
And that's just not the case.
Yeah, because in traditional software, we're used to
going into a configuration file or something like
that and saying, okay, here's the parameters and
you have a discrete value that is going
into this configuration file.
And for that discrete value, you get a
predictable output.

(45:26):
And so you say, if this configuration variable
is X, then I know that I'm going
to get an output that looks like Y.
But with LLMs, you have to, what you're
saying with the prompts is you have to
say, you really don't know what you're gonna
get ever until you test it over and

(45:49):
over again.
You have to say things like, LLM, you
are a chatter creator.
I'm sorry, can I just interrupt?
Yeah, sure.
You're not gonna believe this.
The recording never started.
Oh, crap.
Is somebody recording the stream?
Tell me somebody's recording the stream.

(46:10):
Dave, you have half of it.
You know Cotton Gin's got it.
Cotton Gin, are you recording the stream?
Please tell me someone's recording the stream.
Oh, goodness.
Oh, goodness.
I mean, I've got part of it.
Oh, this is horrible.
When is the last time this happened with
you?
Yeah, Cotton Gin's got it.

(46:31):
Oh, okay, Cotton Gin.
I'm just gonna, so I'm not even gonna
start my, well, I'll start my backup just
in case.
Thank you, Cotton Gin.
Thank you.
I'm sorry.
Dave, back to where you were.
Woo, that was a scary one right there.
No, what I was saying is that you
have to, you're basically, you no longer have

(46:52):
discrete inputs and discrete outputs.
You have suggestions of what outputs may be
based on some vague idea of what the
input may be.
So you have to, like you were saying,
you have to go through and say, model,
you are a chapter creator master for podcasts.

(47:15):
You always are concise.
You have to talk to this thing like
it's a two-year-old.
Or the way I like to look at
it is, you have to talk to it
like it's overseas tech support for your ISP.
Yes, yeah, exactly.
You're not trying to create as many chapters

(47:39):
as you can per minute.
You're trying to be instead very thoughtful about
how many chapters you create because you don't
want to waste people's time.
Like you have to actually talk and you
don't know what the output is going to
be.
It's again, just sort of like, what?
What are we doing?
No, some of the models do have something,

(48:00):
it's called temperature, where you can change how
much variance that it might give you in
a result.
If you, I think it's, I never play
with the temperature, but I think it's if
you put it down to zero, it will,
for the same prompt, it will always give
you the same result for that prompt every
time.
Whereas if you have a higher temperature, then
it might give you something different for the

(48:22):
exact same prompt.
Okay, yeah, see that's, and I think that's
where, I think that's where all of this
becomes so difficult is, and I mean, bless
you for figuring out how to create these
things because like.

(48:43):
Yeah, really.
I think it may just be a personality
thing because I can't, I find interacting with
models hard to tolerate.
Oh yeah, I have that too.
Okay.
I'll tell you a funny story here briefly.
One of the first times I tried using
AI to do something in my code base.
So Podgagement, I built Podgagement completely on my

(49:06):
own.
There was some auto completion for lines of
code, but I wasn't doing any kind of
vibe coding or prompting or anything like that.
But I did want to try and improve
a particular feature.
And so I started trying to use the
AI to improve a feature.
And it was a gauge that I display
in the dashboard that looks kind of like

(49:26):
a speedometer.
It shows the global average rating for a
podcast.
And I was trying to redesign the gauge
because I didn't like the way it looked
with the system I was using.
So the AI gave me a good looking
gauge, but the needle kept going the wrong
direction.
Like 2.5 is halfway between one and
five.
And so it should be straight up.
But for some reason that would be pointing

(49:48):
straight down.
But then 4.5 would be where it's
supposed to be.
And 1.5 would be pointing down below
the thing.
So I was fighting with it, felt like
I was talking with overseas tech support over
and over and over trying to get it
to do it right and sending it screenshots
and all of this stuff.
And then I just decided I need to

(50:09):
just look at the code and figure this
out myself.
And I found the problem was there was
a plus sign where there should have been
a minus sign.
Ugh.
Yeah, yeah, yeah, yeah.
That's classic, classic.
That's, yeah.
I'm just, I think that a product like

(50:31):
PodChapters is, and honestly, the way that AI
is being used in podcasting, just if I
can be, you know, like positive for a
minute, I think that these, what podcasting has

(50:52):
done in general and PodChapters is a great
example is probably the best case scenario for
this stuff.
Like it's a great fit.
I'll take that.
I really can't, yeah.
I can't find any other place in my
life where this stuff fits, but for the
creation of thing where you have like what

(51:13):
you've done where you say, okay, I'm gonna
make a transcript and I'm gonna take this
transcript and I'm gonna feed, you know, and
this transcript is naturally a language corpus and
I'm gonna feed that body into the LLM
and have it do some things to it.
To me, that is like the best case

(51:34):
scenario for where AI is useful.
Exactly.
You know, and I think because you're operating
right in its sweet spot, but you know,
other parts of our lives, I think people
are trying to jam it in and it's
not really gonna fit, but I would say
like with PodChapters, I mean, you're giving these

(51:55):
other formats besides just the 2.0. So
you're giving, I think, PodLove, right?
Yeah.
As well, yeah.
Those kinds of things.
It's like, how are you, so if somebody
is hosting at something like, you know, Buzzsprout,
RSS.com or Blueberry or whatever, if they

(52:16):
take, like, how do they get those things
that they generate on PodChapters into their feed?
It depends on the platform, of course.
And that's one of the things that I'm
trying to document some of these behaviors for
canipodcast.com.
You've probably seen the site caniuse.com, which
is for browsers that shows you like, these
are the CSS things that work in these

(52:37):
browsers as of these versions.
I'm trying to build that for podcasting too,
so you can see exactly which podcast apps
support what features and how they support it.
Like, okay, if they support the description for
the show notes, can you use paragraph tags
in the description?
Can you use headings, bold, italics, hyperlinks?
Does that work in the description?

(52:58):
Stuff like that.
So I'm doing the same thing for hosting
providers and publishing tools to see what do
they support as well.
Do they publish to this tag or do
they support this feature?
How do they support it?
Do they handle the different title tags the
same way?
Like, Buzzsprout, right now, at least, doesn't handle
them the most optimized way, I'll say.

(53:20):
So some of these providers, like Buzzsprout, for
example, if you upload an MP3 that has
the chapters embedded into it, so legacy chapters,
Buzzsprout will use the chapters that are in
the MP3 to then populate their copy of
Podcasting 2.0 and Podlove Simple Chapters.
So they pull that from the file.

(53:40):
A similar thing with rss.com.
Captivate also can pull the chapters from the
MP3, but they don't pull, currently, at least,
they don't pull images and hyperlinks for those
chapters, so you have to go back in
and re-add those.
Or you look at, like, if you're using
PowerPress, I still use PowerPress to generate my
own feed for the Audacity podcast.
And I've made an add-on plugin for

(54:03):
PowerPress that does some more stuff that I
want it to do that are specific to
my workflow.
And inside of PowerPress, since I'm actually still
hosting my media files on Libsyn, but I
don't use Libsyn to create the RSS feed,
I use PowerPress to create the RSS feed.
So in PowerPress, there is a text field
where you can drop in a URL for

(54:24):
chapters, and that's where I now use PodChapters
to host that.
I was previously uploading my chapters to my
own CDN, as well as my transcript, to
my own CDN, and trying to manage that
hosting was another headache.
But so now I use my own product
for that.
I just paste that URL in.
And yeah, it's unfortunate that these different providers

(54:45):
handle it differently, and that's something that I'm
hoping that as PodChapters gets more adoption and
more people use it and have certain questions,
I can bring that to these hosting providers
to say, hey, I've got a customer using
PodChapters, and when they upload their file to
you, you're breaking the chapters, and that's creating

(55:07):
more work for my customer.
It's making them not like you very much.
How about we work together and figure out
a way that you can improve how you're
handling the chapters?
I know at least one hosting company is
in the middle of building an API, and
it would be nice to see hosting companies

(55:28):
develop APIs where you could talk directly.
For external functionality, that's a great idea.
Yeah, that would be really nice to see,
because if you could get an API key
just with your account at your hosting provider
and then plug it into your own different
software tools like PodChapters, and I could think

(55:49):
of 100 use cases right now where that
would be pretty awesome, not just chapters, but
value for value, all kinds of things, where
you could just do things after the fact
on your own system.
That would let you do pretty complex workflows,
like if this, then that type stuff, even

(56:09):
where you don't have like no code type
things.
That's a great idea.
I mean, it's a lot of extra work
for the hosting companies, but man, think of
all, that would actually be cool, all the
extra stuff that could be put on the
outside.
Yeah, and all you'd have to really do
is just wrap it in one of those
like warnings that says, hey, you could really

(56:30):
break your feed, you know, but use it
at your own risk.
But it would be really sweet to be
able to do that.
Because I think, I mean, like even in
your case, I guess you're uploading your Audacity
to Podcast audio to Ellipse and through their
UI?
Yeah, see, I mean, if they had an
API, you could just, you know, pop it

(56:51):
up there.
They do actually have some kind of uploading
API because Auphonic, the audio processing system, you
can publish from Auphonic to certain hosting providers.
So whatever audio you get then from Auphonic
then gets uploaded to the other hosting provider.
And I'm going to work on integrating some
of that stuff for some of the platforms

(57:12):
that support that so that people don't have
to even download the file from PodChapters.
I'd love to see more podcasting focused tools
do that, is integrate with any of the
APIs that are out there so that the
user doesn't have to manage files and download
something that they already uploaded and now they've
got two copies of the same file, which
one was the latest one, but they could

(57:33):
just publish from one system to the other.
Now, do I have to, if I'm going
to use PodChapters to generate something for my
episode, do I have to upload the audio
to you?
Yeah, the audio file.
Okay, because if you could monitor the RSS
feed and have it like publish, if I
could publish to my hosting company, have it

(57:54):
appear in the feed, and then you watch
that feed and then generate stuff based on
my wishes for my latest episode, then I
could just, after the fact, go pull it
off your site and give it to the
hosting provider.
That'd be pretty cool.
Yeah, I might build something like that separate
in the future.
There is already a service that does something

(58:15):
kind of like that, Cast Magic.
Now, they don't do anything in your file
or in the RSS feed or any of
that, but if you wanted social posts generated
from your episode, it can do that, but
this is all after you publish.
I'm a big fan of getting this stuff
together before you publish or, for Adam's sake,
very quickly after you publish.

(58:37):
But there are plenty of tools, yes, that
can monitor your feed and then generate a
transcript or download the file and then suggest
things on top of that, like here are
some social posts you could use to promote
this episode now that it's already published.
So how are you, I'm sorry, go ahead.
No, I was just gonna say, I think
that'd be another area where the API would
be helpful for my hosting provider because I

(58:58):
think most hosting providers have a thing where
you can upload, it's sort of like stage
an episode for release and then publish it
later.
So you're sort of assembling it, getting ready
to go, but it's not in the feed
yet.
That would be a great little way to
integrate and say, oh, there's a sort of
a staged audio here.

(59:20):
I can process that and then you can
go back and do the full publishing later.
Yeah, how are you marketing this?
How do you expect to get it deep
penetration?
Well, so far it is reaching out to
my own audience because the Audacity podcast is
a podcast about podcasting.
So my audience is also my ideal customer.

(59:41):
So I did an episode all about why
you should use chapters and how to use
chapters.
And it flowed very nicely to have native
advertising inside of that episode where I could
promote pod chapters that way.
I have a couple of email lists for
everyone who's ever signed up for the Audacity
podcast email incentives and little promotions that I've
run, as well as podgagement customers, I've promoted

(01:00:03):
it to them.
And then just getting the word out there,
showing it off to other podcasters, trying to
promote it very organically.
I do not do advertising.
I might look into that in the future,
but right now it's just a cashflow thing.
So I'm just trying to grow it organically,
making a product that, here's my perspective on
things, whether it's a podcast or a product,

(01:00:24):
make it so good that you don't have
to advertise it.
Yeah, yeah, yes, that is nice.
I have to say, I've always been impressed
by your entrepreneurial spirit.
Thank you.
You're always coming up with new, how many
different things do you have running now in
your life?
The way I put it is I have

(01:00:45):
my hands in many fires.
Yeah, tell me how many fires.
Tell me about the fires.
I have actual SaaS products too, and I'm
working on a third one that is actually
not podcasting related at all, but I'm partnering
with someone else working on a third SaaS.
So two SaaS's, I have a WordPress plugin,
my podcast about podcasting.
I'm a co-host on the future of

(01:01:07):
podcasting with Dave Jackson.
I do the consulting, I do promotions here
and there for affiliate things.
And he's a dad in his spare time.
And homeschooling nonetheless.
Oh, goodness gracious, wow, wow.
And single too, ladies.

(01:01:29):
The podcasting 2.0 dating services is open
and ready for your business.
We got hot guys, they develop while you
wait.
Well, let's hear some hot namespace talk.
Oh, okay, all right, easy does it now,
easy.
The consulting side of podcasting has, what is

(01:01:51):
the health of that look like now?
I feel like that kind of fell off
a cliff.
Not maybe for you and people who have
been around a very long time and are
established, but for a while there, like from
2018 to 2022 or so, it was like

(01:02:12):
every, you could throw a rock and hit
a podcast consultant.
And now I feel like that is kind
of going away.
Yeah, and it's funny that you made an
unintentional pun, fell off the cliff, because one
of the first major podcast consultants was Cliff
Ravenscraft.
He actually lives not far from me.
We've been long time friends.
Yeah, and I think that now it's not

(01:02:33):
so much only a few people who really
knows how it works, or we don't have
to as an, I almost said the naughty
word that Adam hates, as a world that
podcasts, I'm so hard trying to say not
industry, but as a group of people who

(01:02:56):
work in the podcasting space, they don't have
to know as much now to be successful
as in the early days.
Like before it was all this stuff, like
you had to know how to work with
a mixer.
How do you use an instrument, a thing
that's made for musical instruments and live sound?
How do you use that for podcasting?
And how do you do a mix minus?

(01:03:17):
Now we don't even have to worry about
mix minus.
You just plug in and play things and
it works.
And so the need for the super expertise,
I think has diminished, as well as there
are a lot of people who have either
by luck or by hard work have had
success and they've been able to serve certain

(01:03:38):
companies.
And it's not like there are only a
handful of consultants now that the entire world
is looking to these handful of consultants and
the, what's that cult group?
Not the Gemini, the Illuminati of podcasting doesn't
exist anymore.

(01:03:59):
While there are the people who have stayed
in the industry and are very familiar with
it, I think even just a lot of
companies don't even think necessarily of hiring a
consultant because it's so many people know enough
to get by.
Yeah, I mean, that makes sense.

(01:04:20):
It almost feels like, you remember like from
2000, maybe 2002 to 2006, everybody was a
real estate agent.
Yeah, that turned out real well.
Yeah, you felt like your hairdresser was a
real estate agent.
They were flipping houses on the side.

(01:04:40):
It's like everybody had a real estate license
that you ran into on the street.
And it felt like for a brief period
of time, for a few years, everybody was
a podcast agent or consultant or something like
that.
And you really just don't see any of
that anymore, I don't think, or at least
not to that degree after the great podcast

(01:05:03):
collapse of 2022 or whatever that was.
Has it been that long now?
I guess it's been 22, 23.
And even that collapse is rather misreported because
I've tracked the data, podcastindustryinsights.com is my
other site that has some of this data
that I've been tracking from Apple for a
while.

(01:05:24):
And yes, there was a kind of collapse,
but that's because the fire hose of podcast
pollution from Anchor was cut off.
And so it looked like a collapse of
new podcasts when really it was just the
junk podcasts weren't being put out into the
directories anymore.
So it looked like, but if you were

(01:05:45):
to remove Anchor from the equation, everything looks
like it's still been fairly consistent.
Well, yeah, I think you're talking about raw
numbers.
I was more concerned, I was more referencing
like this, the economics of it, the big
deals and all that.
Oh, right, yeah, the Apple Podcasts update.
Right, yeah, January of 23 is when Spotify

(01:06:08):
started laying people off.
Yes, Nathan would know.
Yes, he would be keenly aware of that
date.
It's burned into his brain.
Speaking of Apple, have we seen any movement
on their possible chapter?
Well, you saw the leak from the-

(01:06:28):
Integration, yeah.
Yeah, where if they put something in the
documentation like that, that means they're gonna support
it.
And I think what probably happened is someone
accidentally pressed publish.
So that's unacceptable, I tell you.
They've been fired, it's not a problem.
Exactly.
Now, here's something that I would love to
bring up with you guys is to actually

(01:06:49):
talk about super chapters because what a chapter
fundamentally is and does has not changed, even
though the way we publish them is so
much better with Podcasting 2.0 and the
way they're published does present other options for
other things you can do with chapters.

(01:07:09):
Still, they are functionally the same, a title,
an image, a link.
How about we make them bigger, make them
better, make them do more?
I think the appetite, this is my hot
take.
I think the appetite for that would probably

(01:07:31):
be bigger if and when Apple does actually
support it.
If that happens, I think you would see
a more desire to start working on that
spec again.

(01:07:56):
Because for the longest time we've heard that,
oh, why are these 2.0 people doing
chapters?
Because it's like, we've already got chapters.
We got chapters in the MP3 file, but
we don't need a whole new chapter spec,
blah, blah, blah.
And to be fair to the Germans, the
Podlove guys got the same thing for years.

(01:08:18):
They got the same type of verbal dismissal
of, well, why are you bothering with that?
We've got MP3 chapters in the ID3 tags.
We don't need all that.
We don't need some new spec and that
kind of thing.
And the benefits that we've been talking about
for so long is that the chapters, you

(01:08:46):
don't put the show notes in the MP3
file.
You don't put the, you know, your album.
Some people do, believe me.
Some people definitely do.
I've seen it.
Sane people do not put the show notes
in their MP3 file.
Like we have all these external repositories for

(01:09:08):
where this structured data goes so that it
doesn't have to live in the audio so
that you don't have to re-upload your
audio every time you want to change something,
which is silly.
So they, you know, the chapters just fits
that same, it fits more, the cloud chapters
fits more along with the model of how

(01:09:28):
the feed already looks.
And I'm, again, mad respect to the Podlove
guys people, but I feel my personal preference,
and maybe I'm biased, I don't know.
A personal sort of idea is that the

(01:09:52):
2.0 chapters fits the feed better than
just listing out the timestamps in the feed
itself.
I don't, I freely admit that I'm not
sure that I could explain why that feels
that way to me right now, but it

(01:10:12):
feels like it fits more of a sort
of a single entity.
So you say like, okay, album art, here's
an image that represents this episode or this
feed, and here's a link to where that
image is.
Right.
And then you say, okay, here's a set
of, here's chapters for this episode, and here's

(01:10:34):
a link to where those are.
Yeah, it's a separation of concerns.
Yes, separation of concern, thank you.
And so like, that is what, that's what
the whole idea has been from the beginning.
So I think that if we can sort
of get, like that's the, what I'm trying
to describe is I think maybe Daniel, that
that's the hurdle to overcome.

(01:10:56):
The hurdle there is to get every, is
to get more people to think this way
where the idea is, okay, we have this
thing called chapters and it lives at this
location and stop thinking about it as something
that is embedded in the media file itself.

(01:11:19):
And if we can make that sort of
mental switch, then I think now, then you
have like a lot of people that are
on the team that are ready to push
forward with adding more stuff to that spec.
Yeah, especially when you start to consider things
like the technological hurdles of supporting even the

(01:11:39):
legacy thing.
Like here's something interesting is that Buzzsprout, although
you can add an image to a podcasting
2.0 chapter on Buzzsprout, they at least
currently do not save that image into the
MP3 file that they host.
So the apps that support the legacy chapters
aren't getting any chapter images, but only the

(01:12:00):
podcasting 2.0 apps are getting the chapter
images.
But even when you think about just chapter
images, if you have a podcast that has
just five images in it, depending on how
those images are saved, you could be significantly
bloating the size of the MP3 file to
embed those images in the file.

(01:12:21):
So it makes total sense for people to
also keep that in mind that do you
really need to be bloating your MP3 file
by embedding all of these high resolution images
in it?
They would be so much better inside of
the external file where then the apps that
support it can load those images.
And then what do you do when somebody

(01:12:41):
like, what's gonna happen to the integrity of
your file also?
Like you load up an MP3 with all
these different, you know, like images and that
kind of stuff and these binary objects, and
then you ship that off to Soundstack to
do dynamic ad insertion, and it comes back
into a CDN.

(01:13:02):
Is that, has that file been molested in
any way?
You have no idea.
I mean, you don't know what's gonna happen
to that file at that point.
Yeah, well, thankfully the MP3 chapter information all
goes in just a specific section of at
least the MP3 file in the headers, but
that does start to mess with certain things.
Like if you're dealing with streaming, all of

(01:13:24):
that header information has to be downloaded before
the audio can even start streaming.
Right, right, yeah.
Yeah, I think if we, I think the
thing you've been calling super chapters for a
long time of just beginning to expand the
capabilities of the chapters file in general, I
think, yeah, I think that is a good

(01:13:46):
idea, but I think we'll probably have, we'll
probably be in a better position to do
that once we get something like Apple to
support it.
Sure, and that is part of the reason
why I built PodChapters is because I thought,
well, when you were like, oh, here it
comes, they're ready, they're good to go, I'm
in, I'm in the pocket.
Well, and I didn't start building it because

(01:14:07):
of Apple.
I actually had started building it already and
then seeing the Apple leak was perfect timing,
but I was thinking that, well, when we
can start making chapters do more than they
do right now, I'd love to make that
possible with my own podcast and love to
start leveraging that.
And so how about I just make a

(01:14:27):
tool that makes it really easy to do
that?
Yeah, yeah, that's the daily source code strategy.
That's exactly right.
We do have to talk about LN versus
a keysend for a moment because we got
a note from Oscar.
Did you get that, Dave?
About the API?
Yeah, whatever it was, my eyes glazed over.

(01:14:51):
I replied to him, I think he's got
what he needs because he was saying that
he needed to be able to put LN
addresses into the shim, but I corresponded with
him and I think he's good.
I think it already supports that.
Well, you want to explain what's going on?
Well, so there's lots to talk about with

(01:15:14):
V4V and we're going to have to talk
about it in depth at some point.
So, but right now, so what Oscar was
saying was that, hey, we're fully supporting LN
address in Fountain, but we need to make

(01:15:39):
sure that the shim, the Podcaster Wallet shim
can also have that data as well.
How do we retroactively implement that?
Well, so podcasterwallet.com just interacts with the

(01:16:01):
podcast index API to insert that stuff in
there.
So podcasterwallet.com will help you create a
value block and then it just talks to
the podcast index API and sets that value
in your feed as a, just a big
JSON object.
And so then it doesn't really care what's

(01:16:23):
the, the podcast index API doesn't really care
what's in that object.
It just takes whatever you give it and
sticks it into the database with that feed.
And so that's, so the part, our quote
unquote partner APIs for like Fountain and Sphinx

(01:16:45):
and those guys.
Sphinx.
Oh, there's a blast from the past.
I know, I know.
But the partner API that I built for
Fountain just allows a feed, allows that value
block to be set.
It just bypasses podcasterwallet.com and just allows
them to set the shim directly.

(01:17:09):
And so he just needs to, he tested
it and it works.
He's just setting LN addresses in the block
now and just encoding them as JSON and
they go straight through.
So it's not a problem.
The only thing that would need to change
would be, I think on our side, I
would need to update the UI for podcasterwallet

(01:17:31):
to be more explicitly accepting of LN address.
Eric said, how many people are using the
shim?
You know, I don't, I think it's very
few, Eric.
I really think, I think there's very few
people that are actually logging into podcasterwallet.com

(01:17:53):
and creating a value block.
Well, it's all API stuff, right?
It's all coming from the APIs, yeah.
So I don't think, honestly, I'm tempted to
get rid of the shim altogether.
That's, a lot of people would be happy
with that.
And now we would still keep the partner

(01:18:13):
APIs, you know, for people, because sometimes these
people just don't have, well, often they don't
have, yeah, I agree, Eric.
I wish you could be retired too.
I may just turn it off and see
what happens.
Because like Fountain, TrueFans, these are gonna, these

(01:18:37):
apps, they need to be able to talk
to the podcast index to be able to
say, hey, here's a value block for this
feed whose hosting company does not support value
blocks.
So that functionality still remains, but the Podcaster
Wallet website probably doesn't even need to exist

(01:19:01):
anymore.
Yes, that's right, Nathan.
Everyone who connects a wallet with Fountain is
technically using the shim, that's correct.
And that's just so that if Fountain or
TrueFans or somebody else sets a wallet on
that feed, it's so that everyone else gets
to see it because it's not in their

(01:19:22):
actual canonical feed.
So that's, and we can't, I don't know,
I still don't know any other way to
do it.
You know, I can't think of another way
to do it.
I will say that I try, our node
is so big.

(01:19:42):
This is where you say, how big is
it?
How big is it?
How big is it?
It's so big.
I've been trying to hook up an LN
address wallet to it, and it's just big,
like everything just- No way.
Everything's, all the voltage stuff just falls down.
It's like, even if you want to run,
what's the dashboard program?

(01:20:07):
Well, it doesn't matter.
None of them work.
Ride the lightning or something?
Yeah, none of that works.
All of it goes, sorry, can't connect now.
It's no good.
And what I don't want to do, I
have an Albi hub, but I don't want
the index's node to be dependent upon my
Albi hub.
So I'm just trying to figure out how

(01:20:27):
to get that done, because we'll switch it.
I've switched pretty much everybody's split over.
Eric PP gave me a really nice list,
so I moved all of those over.
I've had a 1% on my own
Strike wallet for a while.
And it's, I mean, it's pretty good.
The stuff from Fountain comes through with a
link, you know, like the first few words,

(01:20:49):
and then there's a link for the message.
And so that, you know, that link then
takes you to the Boostergram, if there's a
Boostergram.
Strike really sucks because you can't filter it.
So you can't say, show me anything over
one sat, you know, so you have to
scroll through a ton of stuff.
But in principle, it's all kind of working,
which is nice to see.

(01:21:10):
Yeah, I think there's a lot to talk
about with all that.
And again, after finishing this most recent book
that I read, I've got even more, a
more complex set of conflicting thoughts about this
whole thing, you know, about self-sovereignty and

(01:21:33):
how it's so important.
But then also, you know, I don't know.
I'm not maybe fully ready to talk about
it yet.
But I'm hoping to get, hopefully we can
get Chad F on the show.
And because he's been doing a lot of
good work.
He's just been pounding- Non-stop, non

(01:21:54):
-stop.
Non-stop, yeah.
He's just going at it.
Yeah, trying everything, every combination of wallet and
connect and all this stuff.
I mean, it's- It's so valuable, so
valuable.
I've learned so much just watching him try
different configurations of stuff.
And I would say the biggest problem where
we're at right now is it all kind

(01:22:15):
of works, but Helipad doesn't.
You know, that seems to be the big
stumbling block.
Everybody wants Helipad.
Everybody wants to be able to see, you
know, the Boostergrams.
The actual payment part seems to work pretty
well.
Yeah, and it's going to come down to
like, you know, there's,

(01:22:45):
you still have this divide.
You still have this divide between the podcasters
and the listeners, and they have different needs
for the way that, for their setups.
And so that's, you know, Nostra Wallet Connect

(01:23:08):
would work fine for somebody who is a
podcaster.
It's going to be very limiting for somebody
who is a listener, because now they have
to go, it's going to require them to
set up, you know, to set up their
own wallet.
They can't just hook in Strike or Cash

(01:23:31):
App or anything.
And then, but see, my thinking is, I'm
just not, I'm not sure that that's such
a bad thing anymore.
I don't know.
I just need some more time to think
about it.
And hope, and, you know, the, because there's
this thing, like, is it this sort of

(01:23:56):
like growth mindset?
This sort of infatuation with growth.
And in order to get mass adoption for
this technology, we're going to have to do
such and such and such and such.

(01:24:16):
Well, I mean, maybe that's not the right
way to think about it.
Maybe mass adoption is the same thing that
is driving AI off the cliff.
The forever infatuation and obsession with growth and

(01:24:37):
getting scale and these sorts of things, maybe.
That's a very good point.
You know, it may be okay that it
just works in a small way.
That's what I told Tina.
I said, baby, it's okay that it just
works in a small way.
And she married me anyway.

(01:24:59):
There we are.
I've contributed to the show, finally.
I was waiting for my moment.
There we go.
There it is.
We need to thank some people, Dave, because
I got to get you out.
Are you going back to the office?
Are you going to the backyard to work
on the truck?
No, I'm going back to the office.
Okay.
Let me thank, we got a few Boostergrams
during the show.
Martin Lindeskog, he seems to send the same

(01:25:23):
message twice for some reason.
2222.
I like friendly competition.
I've been thinking of getting Frederick Bjorman's app
podcast chapters.
Oh, what is this?
Are you familiar with this, Daniel?
Frederick?
Yeah, that's the same one that we were
talking about.
Oh, we were talking about?
Okay.
He's also a podcaster from Gutenberg, Sweden.
Daniel J.
Lewis, look forward to testing pod chapters.

(01:25:44):
Well, absolutely.
1111 from Hey Citizen, testing keys in a
legacy app from NewLegacyApps.com.
What is that?
NewLegacyApps.com.
All right.
And we got a 10,000 Satoshi boost
from Dreb Scott.
He says, please tell Daniel J.
Lewis that I'm sorry I never got to

(01:26:05):
test this chapters tool.
He did reach out to me.
I wanted to test, but with three small
kids and living my life on the edge
of getting things done, I just couldn't find
the time.
I'm sorry, DJL says, hashtag go podcasting.
You're forgiven.
Yes.
And I think, do we, I don't see
a delimiter this week.
Oh yes, I do.

(01:26:25):
Yes, I do see a delimiter.
Yeah, yeah, yeah, it should be in there.
All right, you're up, Dave.
Oh, let's see.
We've got, pardon me.
I did not have time to print these
out, but so I'm going to read.
I'm just going to, I'm reading straight through.
Go for it.
No prep, Oistein Berha, $5.
He's coming up right after the show, right?
Yes, I believe he is with his mutton
meat and music.

(01:26:47):
Let's see.
Oh, we got an interesting note.
Let's see.
What is that?
It was from the, can't find it.
All right, maybe it'll show up here in
a second.
Kevin Bay, $5.

(01:27:08):
I was looking for a note, because I
saw your response and I was looking for
the note.
Cameron Rose, $25.
Thank you, Cameron.
Appreciate that.
Calaroga Shark Media, LLC sent us $5.
Nice.
It says, been meaning to say thanks.
Thought you'd enjoy my college student homework answers.

(01:27:29):
What's a podcast?
The Revolution Redefined looks at how podcasting began
and explains how Adam Curry and Dave Weiner
used RSS feeds to share audio directly with
listeners.
What's it called?
Revolution Redefined, I think is what he said.
All right, I like that.
Mark Graham, $1.

(01:27:49):
Thank you, Mark.
New Media, that's Martin Linda Skog, $1.
Oh, Brendan over at PodPage, I think they
changed their name, right?
$25.
Isn't PodPage changing their name?
It was Podcast Page changed their name so
that thankfully it's no longer confused with PodPage.
Oh, good, okay.
Thank you, Daniel.
Andrew Grumet, look at him.

(01:28:10):
Oh, Andrew, hello, Andrew Grumet.
It was there in the beginning, no small
contributor.
$100.
Whoa!
Paula, shot caller, 20 inch blades, only in
Paula.
Thank you, brother.
Does he have a note?
Yeah, he says, I can't say it enough.
Thank you guys for the index.
Oh, wow.
You're welcome, Andrew.

(01:28:31):
Thank you, brother.
Joseph Maraca, $5.
Thank you, Joseph.
Oh, Oscar Mary, speak of the devil, $200.
Paula, shot caller, 20 inch blades, only in
Paula.
Paula Ballas, nice.
Thank you all so much.
Basil Phillip, $25.
Thank you, Basil.
Lauren Ball, $24.20, always.

(01:28:53):
PodVerse, that's Mitch and Creon over there, 50
bucks.
Still waiting on PodVerse next gen.
Christopher Harbarick is $10.
Thank you, Christopher.
Mitch Downey's $10.
And that's it.
What happened to that?
Because the use case.
What are you talking about?

(01:29:13):
I'm not sure I understand what you're referring
to here.
Let me see if I can find this
email.
Because we got an email newest.
Got an email from somebody saying how they
had used the API in a really cool
way.
Anyway, that's like my little fly.

(01:29:33):
Mm-hmm.
Oh, oh, wait, that was the, wasn't that
the blind, the accessibility people?
Yeah, yeah, yeah, yeah, yeah.
Yeah, do you remember the details of that
note?
It was more like, I think they create,

(01:29:54):
they have a service for apps that have
accessibility features.
And so they said it was impossible to
get, it was just a nice note.
I mean, it was basically just thanking us,
I think for, this is a few weeks
back.
Oh, here it is, I got it, I
found it.
You found it?
Okay, good.
Yeah, it said, we just discovered your service
and love it.
We are a small Australian company that develops

(01:30:17):
products for the blind and print disabled.
Our main product is a repurposed mobile phone,
which the user talks to and gains access
to.
Right, that's what it was, yeah.
We provide the same, let's see, content on
Alexa speakers.
We have our own small podcast database, which
we find hard to maintain.
We want to use your service, don't want
to put too much load.
Yeah, so they're using that for, using us

(01:30:38):
for that and that's great.
Love that.
I love it.
Love it.
Yeah, that's fantastic.
Did they send a donation?
No, they forgot that part, didn't they?
Nice, nice guys.
Enjoy the service while it lasts.
Let's see, Anonymous 5170, that's true fan support.

(01:31:00):
Oh, thank you.
From Anonymous, got a few of those.
Bruce the Ugly Whacking Duck sent 2222 through
Podcast Guru, crazy ride on this episode from
Hackers to Keysend.
Glad you gave us a chance to hear
it.
It might be a Hackers 2 movie, 73.
Yeah, I don't think so, I don't think
so.

(01:31:20):
Oh, look at Dave Jackson, 9861.
That's your partner in crime, Daniel.
Nice.
Nice.
Through Castamatic, he says 9861, the Todd Cochran
boost.
9861.
He died on September 9th and he was
61 years old.
9861, forever now enshrined as the Todd boost.

(01:31:44):
The Todd boost.
Rest in peace, brother.
Oh, Eric from R Podcasting, 190,000 sats.
What?
Baller, shot caller, 20 inch blades on the
Impala.
Whoa, that's a nice boost.
Yikes, hi Dave and Adam.

(01:32:05):
I know I've been away for most of
the year but I'm slowly catching back up.
This might be the first ever apology boost.
Not the first.
Because I saw that the GitHub actions to
update the fancy podcast index database dashboard analysis
had stopped running for almost a year.
Ah, yes.
GitHub disables scheduled actions if there are no

(01:32:25):
commits for more than 60 days.
You know, I did not know that.
Oh, really?
Oh, that's important information.
It's good to know.
I fixed that problem and now the dashboard
has the latest and greatest analyses of duplicates
and quality checks.
I'll do a better job of monitoring, I
promise.
And go podcasting.
Okay, let's go podcasting.
Go podcasting!
I'm going to throw out a go podcasting

(01:32:46):
for that boost.
Apology boosts are the best, aren't they?
All right, I just popped this, yes.
I just popped this link from Eric over
here into the boardroom.
Thank you, brother.
Appreciate that, man.
Oh, Kyron from Mirimodal's podcast.
One, one, one, one.
Essential Richard from Fountain.
He says, we demand change around here.
More beers in the boardroom.

(01:33:06):
Okay.
There's your change.
No beer in the boardroom allowed, Kyron.
No, no, no.
Donuts, yes.
Beer, no.
Yep.
And whiskey.
And the Delimiter Comma Street Blogger, 12,450
sats through Fountain.
He says, howdy, Dave and Adam.
Bros, for your female partners, girlfriends, hoes, or

(01:33:30):
wives, I want to recommend a podcast by
ladies for ladies entitled Dumpster Fire with Bridget
Fatesi.
Quote, Dumpster Fire is a satirical show created
and hosted by comedian Bridget Fatesi.
I think it's Fatasi.
Fatasi.

(01:33:51):
That takes a look at the weekend news,
current events, and politics.
We make burgers out of your sacred cows.
Delicious, end quote.
It's in podcastindex.org or www.fatasi.com.
Yo, CSB, the maker of www.trading.toys.
It's like the third person who's reached out
to me about her.

(01:34:12):
Someone connected me with her.
Someone from Glenn Beck Show said, I know
I'm overstepping social boundaries, but Bridget's a sister
in Christ and I want to make sure
you people in the hill country knew about
each other.
She lives in Georgetown, which is 100 miles
away, but okay.
And then someone else sent me a note
and said you should listen to what Bridget's

(01:34:33):
doing and help her with value for value.
So three times, after three times, that means
I got to do something about it.
So I'll reach out to her.
Speaking of hill country, I found out the
other day my wife teaches at a Charlotte
Mason school, if you know anything about the
Charlotte Mason method.

(01:34:55):
Yeah, but there's a name here, not heritage.
Ambleside.
Ambleside, yes, Ambleside, right down the road here.
That's right, yep.
And so my wife started teaching Charlotte Mason
educational philosophy, gosh, many, many years ago.
Probably 15 years ago.
And she used Ambleside online to do that

(01:35:16):
with our kids.
And now she teaches at a Charlotte Mason
school.
And I just found out the other day
that the people that created Ambleside online are
from Fredericksburg.
Yeah, that's right.
I've actually, they, this is one of those,
hey, come on, talk to us about podcasting,
because some people go to church, they work
at Ambleside.
So I met with everybody, met with all

(01:35:36):
the leadership.
And after that, pretty much crickets, that's how
it goes.
Yeah, but that's a huge, I mean, that's
a huge deal in that world.
Well, we're trying to get them to do
some podcasting.
What always happens is the leadership wants to
do the podcast.
I'm like, but you need to get the
kids podcasting.

(01:35:57):
You don't understand this is where it needs
to go.
I'm like, oh, yeah, I don't really think
about that.
Hey, Daniel J.
Lewis, man, thank you so much for being
here.
Congratulations with the launch of PodChapters.
I'll be praying for Apple to add chapters
soon, and that it'll fit seamlessly, that they
should create an API for all your stuff
to just flow in.

(01:36:18):
Thank you.
And thank you to everybody who supported us
here at podcastindex.org.
Of course, everything goes into the fund to
keep the index running.
You can go to podcastindex.org down at
the bottom.
There's a big red donut button.
There's a donut button there.
A big red, I want that button.
Hit the donut button and support us with

(01:36:40):
your Fiat fund coupons.
We always accept the boost, of course.
We really appreciate that.
Brother Dave, I hope you feel better.
And you take it easy on the truck
in the backyard.
Don't be so mad at the AI.
Daniel, any parting words?
Go podcasting.
There it is, everybody.
Thank you very much, boardroom.

(01:37:00):
Cotton Gin, please give me the link to
the recording so we can actually post this.
We'll see you next week here on Podcasting
2.0. Podcasts

(01:37:21):
are cool.
Do you hope you have been listening to
Podcasting 2.0?
Visit podcastindex.org for more information.
Go podcasting!
Player A enters the game.

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Dateline NBC

The Burden

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Episode 237: FF-P-MEG

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Dateline NBC

The Burden

All Episodes

Episode 237: FF-P-MEG

Stuff You Should Know